Dynamic Data Structures. John Ross Wallrabenstein

Transcription

1 Dynamic Data Structures John Ross Wallrabenstein

2 Recursion Review In the previous lecture, I asked you to find a recursive definition for the following problem: Given: A base10 number X Assume X has N digits Output: X printed vertically

3 Print Vertically Did anyone come up with a solution?

4 Print Vertically

5 Print Vertically What is our base case?

6 Print Vertically What is our base case? When X is a single digit: print X

7 Print Vertically What is our base case? When X is a single digit: print X What is our recursive call?

8 Print Vertically What is our base case? When X is a single digit: print X What is our recursive call? When X has more than one digit:

9 Print Vertically What is our base case? When X is a single digit: print X What is our recursive call? When X has more than one digit: Print the N-1 most significant digits

10 Print Vertically What is our base case? When X is a single digit: print X What is our recursive call? When X has more than one digit: Print the N-1 most significant digits Print the least significant digit

11 Print Vertically public class PrintVertically { public static void main(string[] args) { } int x = 1234; System.out.println("writeVertical(" + x + "):"); writevertical(x); public static void writevertical(int x){ //If X is a single digit if(x < 10){ System.out.println(x); } //If X has more than a single digit else{ //Print the most significant digits vertically writevertical(x/10); } } } //Print the least significant digit vertically System.out.println(x%10);

12 Asymptotic Notation When analyzing an algorithm, it is convenient to have a simple way to express the efficiency of the algorithm This allows us to compare the efficiency of two different algorithms

13 Big-Oh Notation Assumption: Our algorithm is defined in terms of functions whose domains are the set of natural numbers N={0,1,2,...}

14 Big-Oh Notation We would like to bound the running time of our algorithm by a measure of the number of atomic (constant time) operations it performs

15 Big-Oh Notation We say that a function f(n) is O(g(n)) iff: Not a typo! there exists some constant c and n0 such that:

16 Big-Oh Notation Formally:

17 Big-Oh Notation

18 Big-Oh Notation Example: This program runs in O(n) time for(int i = 0; i < n; ++i){ //Atomic Operations }

19 Big-Oh Notation Example: This program runs in O(n 2 ) time for(int i = 0; i < n; ++i){ //Atomic Operations for(int j = 0; j < n; ++j){ //Atomic Operations } }

20 Asymptotic Analysis The smaller the asymptotic bound, the faster the algorithm e.g. O(n) is preferable to O(n 2 )

21 Data Structures

22 Data Structures What are they?

23 Data Structures What are they? A data structure is an abstraction used to model a set of related data

24 Data Structures What are they? A data structure is an abstraction used to model a set of related data What data structures are you familiar with?

25 Data Structures What are they? A data structure is an abstraction used to model a set of related data What data structures are you familiar with? Most Common Example: Arrays

26 Data Structures What are they? A data structure is an abstraction used to model a set of related data What data structures are you familiar with? Most Common Example: Arrays Food for Thought: File Systems

27 Static Data Structures Static Data Structures Data Structures that do not change* over the lifetime of their use *This usually refers to the size of the structure

28 Static Data Structures Benefits: Fast Access to Individual Elements Drawbacks: Expensive to update e.g. Insertion, Deletion Finite size known at compile time e.g. Arrays

29 Dynamic Data Structures Dynamic Data Structures Data Structures that change over their lifetime, allowing for efficient updates

30 Dynamic Data Structures Benefits: Fast Updates (Insertion, Deletion) Flexible Size Drawbacks: Slow access to individual elements

31 Arrays Recall that the size of an array must be declared at compile time How does this affect insertion/deletion?

32 Array Deletion When we delete an element from an array, we usually waste memory If we want to reclaim this memory, we have to create a new array (of size n-1) and copy all n-1 elements This is inefficient!

33 Array Insertion Case: Insert an element between existing elements We have to copy the entire array to a new location, leaving space for the new element This is inefficient!

34 Array Insertion Case: Insert (push) an element onto the end of a full array We have to: Allocate space for an array of size n+1 Copy n elements into the new array Add new element to the end Deallocate memory for old array (For C Programmers)

35 Array Insertion Can we do better than O(n) time? How?

36 Array Insertion Improving Insertion Efficiency Rule: When the array is full and a new element is to be added, proceed by: Allocating space for an array of size 2n Copy original n elements Add new element

37 Array Insertion How does this help? In the amortized case, we now have that insertion takes O(1)! Amortized analysis involves taking the cost of an operation over the number of times the operation is invoked In this case: A single push is effectively O(1)

38 ArrayLists You have probably seen ArrayLists An ArrayList is a hybrid between a static data structure (arrays) and a fully dynamic structure ArrayLists use the previous optimization when inserting elements into a full array

39 Linked Lists A linked list is a dynamic data structure composed of nodes Each node contains: Data Pointer to next node in list* *In the case of a singly linked list

40 Linked Lists How is the order of the elements in an array determined? By the index of the element How is the order of the elements in a linked list determined? By the next pointer in each node

41 Linked Lists What are the benefits of this structure? Insertion: Point the current node s next to predecessor s next Point predecessor s next to the current node No resizing!

42 Linked Lists What are the benefits of this structure? Deletion: Point the previous node s (predecessor) next to the current node s next (successor) No inefficient use of memory!

43 Linked Lists What are the drawbacks of this structure? Representation on disk may be fragmented Why? Linked Lists do not need contiguous blocks of memory Thus, nodes can be stored anywhere

44 Dynamic Sets When building algorithms, sets of data must be able to change over time* We must be able to: Insert Delete Test Membership *Have we seen this phrase before?...

45 Dynamic Sets What dynamic sets have you seen before? Stacks Queues Linked Lists Etc... Let s look at a particular application of sets

46 Disjoint Sets Two sets A and B are said to be disjoint if their intersection is null Many times, it is useful to partition a larger set into smaller, disjoint sets How can we keep track (Find) of the elements and combine (Union) the smaller sets into larger sets efficiently? Dynamic Data Structures to the rescue!

47 Disjoint Sets A disjoint-set data structure maintains a collection S={S1,S2,...,Sk} of disjoint dynamic sets Each set is identified by a representative

48 Disjoint Set Operations We require the following operations to be supported: Make-Set(x) - create a new set whose only member is x Union(x,y) - unites the dynamic sets that contain elements x and y Find-Set(x) - return a pointer to the representative of the set containing x

49 Why We Care Before proceeding, why should we worry about how to use dynamic data structures to handle dynamic disjoint sets? Common Application: Determining the Connected Components of a Graph See Board

50 Union / Find We will focus on constructing a dynamic data structure that computes Union and Find efficiently

51 Union / Find To support the operations Union and Find on disjoint sets, we will use a disjoint set forest of trees Setup: Each member points only to its parent The root is the representative of the set (and also its own parent)

52 Union / Find

53 Find To implement Find(x) and return its representative element: Follow parent pointers until the root is reached Return the root (representative)

54 Union To implement Union(x,y) Set the parent of one of the set s representatives to the root of the other set

55 Union / Find Consider the following scenario: We have x1,x2,...,xn objects We make each xi a set with Make-Set(x) We Union each xi into one large set See Board Does order matter?

56 Union / Find Consider the running time of Find if we only use the tools we have seen so far: To Find element X in a set, if we have a linear chain of n nodes takes O(n) time If we have n calls to Find our running time is then: Can we do better?

57 Union / Find What is the source of inefficiency? We are appending a larger set onto a smaller set (xi) How can we fix this? Maintain the size of each set (easy) Always append the smaller set onto the larger set Use path compression to reduce tree height

58 Union By Rank To always append the smaller set to the larger set, we maintain a rank for each set The rank is an upper bound on the height of the set tree The root with the smaller rank points to the root of the larger set

59 Union By Rank See Board for Example Performance Analysis: Where we perform m operations on n elements The proof of this bound is beyond the scope of this course

60 Path Compression To lower the number of parent nodes we have to pass through to reach the root in a Find(x) operation: Run Find(x) once to retrieve the root Run Find(x) again, pointing each node along the way to root See Board Running time:

61 Combining Both When both Union By Rank and Path Compression are used: Running Time: Here, alpha represents the Ackermann Function, which is effectively constant

62 Questions